release: v0.4.10

CHANGELOG.md

@@ -7,6 +7,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.4.10] - 2026-08-22
+
 ### Fixed
 - Absolute uinput pointer axes now end at the final logical desktop pixel, so
   edge coordinates are advertised and clamped consistently.
@@ -454,7 +456,8 @@ pages; also bumps the MCP server's advertised version string to match.
 - Validated against GNOME 50.1 on Wayland (Ubuntu 25.10).
 - KDE / Sway / Hyprland untested — see README support matrix.
 
-[Unreleased]: https://github.com/agent-sh/computer-use-linux/compare/v0.4.9...HEAD
+[Unreleased]: https://github.com/agent-sh/computer-use-linux/compare/v0.4.10...HEAD
+[0.4.10]: https://github.com/agent-sh/computer-use-linux/compare/v0.4.9...v0.4.10
 [0.4.9]: https://github.com/agent-sh/computer-use-linux/compare/v0.4.8...v0.4.9
 [0.4.8]: https://github.com/agent-sh/computer-use-linux/compare/v0.4.7...v0.4.8
 [0.4.7]: https://github.com/agent-sh/computer-use-linux/compare/v0.4.6...v0.4.7

Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "computer-use-linux"
-version = "0.4.9"
+version = "0.4.10"
 edition = "2021"
 authors = ["Avi Fenesh <avifenesh@users.noreply.github.com>"]
 license = "MIT"

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@agent-sh/computer-use-linux",
-  "version": "0.4.9",
+  "version": "0.4.10",
   "description": "Linux desktop-control MCP server: AT-SPI accessibility trees, Wayland/X11 input, screenshots, and compositor window targeting.",
   "license": "MIT",
   "type": "commonjs",

src/diagnostics.rs

@@ -463,9 +463,9 @@ fn desktop_env_hydration_updates(
         && current_env
             .get("DISPLAY")
             .is_some_and(|value| !value.trim().is_empty())
-        && !current_env
+        && current_env
             .get("WAYLAND_DISPLAY")
-            .is_some_and(|value| !value.trim().is_empty());
+            .is_none_or(|value| value.trim().is_empty());
 
     DESKTOP_ENV_KEYS
         .iter()

src/server.rs

@@ -1688,7 +1688,7 @@ impl ComputerUseLinux {
     // The rmcp tool_handler macro only accepts a string literal here, so this
     // can't be env!("CARGO_PKG_VERSION"); the MCP safety check (CI) fails the
     // build if it drifts from the Cargo version.
-    version = "0.4.9",
+    version = "0.4.10",
     instructions = "Begin every turn that uses Computer Use by calling get_app_state. If diagnostics report disabled GNOME accessibility, call setup_accessibility before asking the user to retry. Use list_windows/focused_window before targeted keyboard input. If diagnostics report windowing.can_list_windows=false on GNOME, call setup_window_targeting to install the optional GNOME Shell extension backend, then ask the user to log out and back in if the setup report says a shell reload is required. This Linux backend can capture size-bounded screenshots through GNOME Shell or XDG Desktop Portal, read AT-SPI trees with action/value metadata, invoke native AT-SPI actions, set AT-SPI values or editable text, list/focus compositor windows through registered Linux window backends when the session permits it, attach best-effort terminal tty/process metadata to terminal windows, send coordinate or element-targeted click/scroll/drag input through the Wayland remote desktop portal when available, and send layout-safe literal type_text through KDE clipboard integration on Plasma Wayland or through portal keysyms on other Wayland sessions before falling back to ydotool. Screenshot results include width/height for the returned image plus coordinate_width/coordinate_height and scale for desktop coordinate conversion; request more detail with max_width, max_height, max_bytes, format=jpeg, quality, or a smaller target/crop instead of relying on unbounded screenshots. Tools with readOnlyHint=false may mutate local desktop or application state; hosts should require approval for actions that can submit, delete, send, purchase, or overwrite data. For element-targeted actions, prefer element_index from the latest get_app_state result; click, perform_action, and set_value can also use semantic role/name/text/states selectors when the target is unique. type_text and press_key accept optional window_id, pid, app_id, wm_class, title, tty, terminal_pid, terminal_command, or terminal_cwd selectors and refuse targeted input if focus cannot be verified. After targeted keyboard input, results append focused-element feedback from AT-SPI (role, name, editable) and warn when no editable element holds focus — treat that warning as the input not landing. Screenshot, click, and input results warn when the target window or coordinate is partially or fully off-screen; use move_window/resize_window (GNOME Shell extension backend) to bring a window fully on-screen before retrying. scroll accepts the same window targeting and relative coordinates as click. get_app_state returns a compact readiness block by default; pass verbose=true for the full diagnostics dump. Electron apps expose no AT-SPI tree unless launched with --force-renderer-accessibility."
 )]
 impl ServerHandler for ComputerUseLinux {}
@@ -3612,7 +3612,7 @@ fn native_x11_xdotool_pointer_session(
     wayland_display: Option<&str>,
 ) -> bool {
     session_type.is_some_and(|value| value.trim().eq_ignore_ascii_case("x11"))
-        && !wayland_display.is_some_and(|value| !value.trim().is_empty())
+        && wayland_display.is_none_or(|value| value.trim().is_empty())
 }
 
 fn prefer_xdotool_pointer(